From a4720b44cda1754cac3cb57b21fc045131a46401 Mon Sep 17 00:00:00 2001
From: poire-z <poire-z@users.noreply.github.com>
Date: Mon, 3 Jul 2023 19:49:22 +0200
Subject: [PATCH] Text search: various Kopt search fixes

- Properly parse input text for words (the previous
  code wasn't working with Greek letters)
- With multiple words search, don't allow "substring
  matching" for words in the middle
- Remove support for Lua pattens, so to get proper
  substring matching (as we have with cre text search)
---
 frontend/document/koptinterface.lua | 42 ++++++++++++++++++++++-------
 spec/unit/readersearch_spec.lua     |  6 -----
 2 files changed, 33 insertions(+), 15 deletions(-)

diff --git a/frontend/document/koptinterface.lua b/frontend/document/koptinterface.lua
index b8be38da5..e5fef4ac8 100644
--- a/frontend/document/koptinterface.lua
+++ b/frontend/document/koptinterface.lua
@@ -13,6 +13,7 @@ local Geom = require("ui/geometry")
 local KOPTContext = require("ffi/koptcontext")
 local Persist = require("persist")
 local TileCacheItem = require("document/tilecacheitem")
+local Utf8Proc = require("ffi/utf8proc")
 local logger = require("logger")
 local util = require("util")
 
@@ -1337,18 +1338,22 @@ end
 local function all_matches(boxes, pattern, caseInsensitive)
     -- pattern list of single words
     local plist = {}
-    -- split utf-8 characters
-    for words in pattern:gmatch("[\32-\127\192-\255]+[\128-\191]*") do
-        -- split space seperated words
-        for word in words:gmatch("[^%s]+") do
-            table.insert(plist, caseInsensitive and word:lower() or word)
+    -- (as in util.splitToWords(), but only splitting on spaces, keeping punctuations)
+    for word in util.gsplit(pattern, "%s+") do
+        if util.hasCJKChar(word) then
+            for char in util.gsplit(word, "[\192-\255][\128-\191]+", true) do
+                table.insert(plist, caseInsensitive and Utf8Proc.lowercase(util.fixUtf8(char, "?")) or char)
+            end
+        else
+            table.insert(plist, caseInsensitive and Utf8Proc.lowercase(util.fixUtf8(word, "?")) or word)
         end
     end
+    local pnb = #plist
     -- return mached word indices from index i, j
     local function match(i, j)
         local pindex = 1
         local matched_indices = {}
-        if #plist == 0 then return end
+        if pnb == 0 then return end
         while true do
             if #boxes[i] < j then
                 j = j - #boxes[i]
@@ -1356,10 +1361,27 @@ local function all_matches(boxes, pattern, caseInsensitive)
             end
             if i > #boxes then break end
             local box = boxes[i][j]
-            local word = caseInsensitive and box.word:lower() or box.word
-            if word:match(plist[pindex]) then
+            local word = caseInsensitive and Utf8Proc.lowercase(util.fixUtf8(box.word, "?")) or box.word
+            local pword = plist[pindex]
+            local matched
+            if pnb == 1 then -- single word in plist
+                matched = word:find(pword, 1, true)
+            else -- multiple words in plist
+                if pindex == 1 then
+                    -- first word of query should match at end of a word from the document
+                   matched = word:sub(-#pword) == pword
+                elseif pindex == pnb then
+                    -- last word of query should match at start of the word from the document
+                    matched = word:sub(1, #pword) == pword
+                else
+                    -- middle words in query should match exactly the word from the document
+                    matched = word == pword
+                end
+            end
+            if matched then
                 table.insert(matched_indices, {i, j})
-                if pindex == #plist then
+                if pindex == pnb then
+                    -- all words in plist iterated, all matched
                     return matched_indices
                 else
                     j = j + 1
@@ -1370,6 +1392,8 @@ local function all_matches(boxes, pattern, caseInsensitive)
             end
         end
     end
+    -- Note that this returns a full word box, even if what matches
+    -- is only a substring of a word box.
     return coroutine.wrap(function()
         for i, line in ipairs(boxes) do
             for j, box in ipairs(line) do
diff --git a/spec/unit/readersearch_spec.lua b/spec/unit/readersearch_spec.lua
index dedbfbb9a..bf267e68f 100644
--- a/spec/unit/readersearch_spec.lua
+++ b/spec/unit/readersearch_spec.lua
@@ -151,12 +151,6 @@ describe("Readersearch module", function()
         it("should match whole phrase in one page", function()
             assert.are.equal(1*3, #doc.koptinterface:findAllMatches(doc, "mean that the", true, 20))
         end)
-        it("should match with lua pattern", function()
-            assert.are.equal(7*1, #doc.koptinterface:findAllMatches(doc, "chapter", true, 30))
-            assert.are.equal(3*2, #doc.koptinterface:findAllMatches(doc, "chapter %d", true, 30))
-            assert.are.equal(2*2, #doc.koptinterface:findAllMatches(doc, "chapter %d%d", true, 30))
-            assert.are.equal(0*2, #doc.koptinterface:findAllMatches(doc, "chapter %d%d%d", true, 30))
-        end)
         it("should not match empty string", function()
             assert.are.equal(0, #doc.koptinterface:findAllMatches(doc, "", true, 1))
         end)